What is the bootstrap method?

The Bootstrap method is a very useful statistical estimation method. Bradley Efron, a professor in the Department of Statistics at Stanford, proposed a new method based on summarizing and summarizing previous research results. non-parametric statistical methods.

Bootstrap is a type of non-parametric Monte Carlo method. Its essence is to resample the observation information and then make statistical inferences about the distribution characteristics of the population.

Because this method makes full use of the given observation information, it does not require other assumptions of the model and the addition of new observations, and it is robust and efficient. Since the 1980s, with the introduction of computer technology into statistical practice, this method has become more and more popular and is widely used in the field of machine learning.

First of all, Bootstrap can avoid the sample reduction problem caused by Cross-Validation through resampling. Secondly, Bootstrap can also be used to create randomness in the data. For example, the first step of the well-known random forest algorithm is to randomly select k new bootstrap sample sets with replacement using the bootstrap method from the original training data set, and thereby construct k classification regression trees.

The above is the detailed content of What is the bootstrap method?. For more information, please follow other related articles on the PHP Chinese website!